Is Generalised Cost Justified?

Is Generalised Cost Justified?


M Wardman, ITS, University of Leeds, UK


The concept of Generalised Cost is widely used in transport planning. This paper reports tests of whether the cross-sectional and inter-temporal variations in elasticities implied by this approach are empirically justified.


Standard transport planning makes much use of the concept of generalised cost (GC), whereby the time related aspects of travel are combined with the monetary cost through the use of the appropriate values of time. The elasticities to the separate components depend upon the proportion that they form of GC, although without any explicit testing of whether this is empirically justified. A significant implication of this approach is that, with monetary values increasing over time with income, the time related elasticities will tend to increase over time and the implied fare elasticity will fall, all other things equal.

In contrast, what might be regarded to be a conventional economics ?textbook? approach would specify separate elasticities to each of the attributes regarded to influence travel. The disaggregate modelling approach essentially estimates elasticites to each attribute component of the utility function.

Yet another approach, although seemingly unique to the rail industry in Great Britain, is to merge the time related aspects into a single variable, termed generalised journey time (GJT) but to maintain fare as a separate term. This approach also makes certain assumptions of the pattern of implied elasticity variation, although less so than the GC approach.

In Britain, we are blessed with large amounts of data relating to rail travel demand between stations across many years. Whilst the cost of rail travel does exhibit considerable variation over routes and time, in line with other travel modes, what is unique to the railway industry is that some routes also exhibit considerable variations in the journey time, service frequency, interchange and other components of service quality. This supports extensive modelling opportunities not otherwise possible.

This paper exploits the availability of very large data sets relating to different types of flow and exhibiting variations in both fare and timetable related service quality variables in order to:

? provide fresh empirical evidence as to the GC elasticity since surprisingly little such evidence exists for such a widely used parameter;

? determine whether the GC approach of standard transport planning provides a better explanation of rail travel than the GJT and fare approach or an approach that specifies separate elasticities to each component of GC;

? test whether the spatial and temporal variations in elasticities to constituent variables implied by the use of composite terms such as GC and GJT are empirically justified;

? determine the appropriate values of time to use conditional upon the continued use of the standard GC approach;

We here present results from a number of very large data sets of rail travel between stations. Models are estimated using the GC approach, the GJT and fare approach and the separate elasticities to each variable approach, allowing the value of time to be freely estimated rather than imposed from other evidence.

Not surprisingly, the results vary by type of flow and across data sets, but the general conclusion is quite clear in that the GC approach does not generally provide the best fit to the data.

We have tested the implied elasticity variation for the component variables of GC and found that the variation implied by the GC approach cannot be empirically justified cross-sectionally in comparison to freely estimated elasticities. Models which estimate elasticities to the separate components of GC perform well.

Nor is there evidence to support variation in elasticities over time in line with the GC formulation. In addition, the paper reports some challenging findings regarding the appropriate values of time to use in constructing GC given that this approach is used.


Association for European Transport